ai-browser-profile 1.0.8 → 1.0.10

package/ai_browser_profile/cookies.py

@@ -143,6 +143,22 @@ def read_cookies(
         )
 
     domain_filters = list(domains) if domains else None
+
+    def _host_matches(host: str) -> bool:
+        # Domain-suffix match: 'x.com' matches 'x.com' / 'api.x.com' but not 'fedex.com'.
+        # Cookie host_keys often start with '.' for "all subdomains" — strip that.
+        h = host or ""
+        if "://" in h:
+            h = h.split("://", 1)[1]
+        h = h.split("/", 1)[0].split(":", 1)[0].lstrip(".").lower()
+        for f in (domain_filters or []):
+            ff = (f or "").strip().lstrip(".").lower()
+            if not ff:
+                continue
+            if h == ff or h.endswith("." + ff):
+                return True
+        return False
+
     key = _derive_key(_keychain_password(profile.browser))
     cookies: list[Cookie] = []
     skipped = 0
@@ -167,7 +183,7 @@ def read_cookies(
         )
         for row in rows:
             host = _txt(row["host_key"])
-            if domain_filters and not any(d in host for d in domain_filters):
+            if domain_filters and not _host_matches(host):
                 continue
             value = _txt(row["value"])
             if not value and row["encrypted_value"]:

package/ai_browser_profile/indexeddb.py

@@ -179,15 +179,65 @@ def read_indexeddb(
         [o.strip() for o in origins if o and o.strip()] if origins else None
     )
 
+    def _host_matches(origin: str) -> bool:
+        # Domain-suffix match: filter 'x.com' matches 'x.com' and
+        # 'api.x.com' but NOT 'fedex.com' / 'swiftpackageindex.com'.
+        h = origin or ""
+        if "://" in h:
+            h = h.split("://", 1)[1]
+        h = h.split("/", 1)[0].split(":", 1)[0].lstrip(".").lower()
+        for f in (origin_filter or []):
+            ff = (f or "").strip().lstrip(".").lower()
+            if not ff:
+                continue
+            if h == ff or h.endswith("." + ff):
+                return True
+        return False
+
+    # Defaults to skip even when no explicit filter is given:
+    #   chrome-extension://  — extensions, not portable across browsers
+    #   localhost / 127.*    — dev servers, irrelevant across machines
+    #   file://              — local file URLs
+    SKIP_PREFIXES = (
+        "chrome-extension://",
+        "http://localhost",
+        "https://localhost",
+        "http://127.",
+        "https://127.",
+        "file://",
+    )
+    # Skip pathologically large origins by default (e.g. kapwing video editor
+    # which stores 2 GB of project blobs). Caller can still ask for them
+    # explicitly via origin_filter.
+    MAX_LEVELDB_BYTES = 200 * 1024 * 1024  # 200 MB
+
+    def _dir_size(p) -> int:
+        try:
+            return sum(f.stat().st_size for f in p.rglob("*") if f.is_file())
+        except Exception:
+            return 0
+
     out: dict[str, list[IdbDbDump]] = {}
     skipped_dbs = 0
+    skipped_origins = 0
 
     for leveldb_dir in sorted(idb_root.glob("*.indexeddb.leveldb")):
         origin = _idb_dir_to_origin(leveldb_dir.name)
         if origin is None:
             continue
-        if origin_filter and not any(f in origin for f in origin_filter):
-            continue
+        if origin_filter:
+            if not _host_matches(origin):
+                continue
+        else:
+            # No explicit filter — apply default safety skips.
+            if any(origin.startswith(p) for p in SKIP_PREFIXES):
+                skipped_origins += 1
+                continue
+            size = _dir_size(leveldb_dir)
+            if size > MAX_LEVELDB_BYTES:
+                log.info("skipping oversized IndexedDB %s (%.1f MB)", origin, size/1024/1024)
+                skipped_origins += 1
+                continue
 
         blob_dir = leveldb_dir.parent / leveldb_dir.name.replace(".leveldb", ".blob")
 

package/ai_browser_profile/localstorage.py

@@ -33,6 +33,29 @@ from ai_browser_profile.cookies import _ws_from_cdp_url, find_profile
 log = logging.getLogger(__name__)
 
 
+def _host_matches(host: str, filters: list[str]) -> bool:
+    """Domain-suffix match: filter 'x.com' matches 'x.com' and 'sub.x.com',
+    but NOT 'fedex.com' or 'swiftpackageindex.com'.
+
+    Accepts either a raw host like 'example.com' or a full origin like
+    'https://example.com' / 'https://example.com:8080'.
+    """
+    if not host:
+        return False
+    h = host
+    if "://" in h:
+        h = h.split("://", 1)[1]
+    h = h.split("/", 1)[0].split(":", 1)[0]  # strip path and port
+    h = h.lstrip(".").lower()
+    for f in filters:
+        ff = f.strip().lstrip(".").lower()
+        if not ff:
+            continue
+        if h == ff or h.endswith("." + ff):
+            return True
+    return False
+
+
 def read_localstorage(
     profile: BrowserProfile,
     origins: Optional[Iterable[str]] = None,
@@ -77,7 +100,7 @@ def read_localstorage(
                 value = record.value
                 if not origin or not key or value is None:
                     continue
-                if origin_filters and not any(f in origin for f in origin_filters):
+                if origin_filters and not _host_matches(origin, origin_filters):
                     continue
                 if isinstance(value, bytes):
                     try:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-browser-profile",
-  "version": "1.0.8",
+  "version": "1.0.10",
   "description": "Extract user identity (name, emails, accounts, addresses, payments) from browser data into a self-ranking SQLite database. Install as a Claude Code agent skill.",
   "bin": {
     "ai-browser-profile": "bin/cli.js"